amilios

12 ms·

121.
by amilios 4y ago
Compare housing costs in the 1970s vs now.
122.
by amilios 4y ago
(Seemingly) Inevitable climate change and ecological destruction, growing income inequality and falling standards of living, a deathly pandemic that's still killing people while we pretend everything is back to normal (+ long covid pro
123.
by amilios 4y ago
eating disorder
124.
by amilios 4y ago
Comparing the 13B model here https://huggingface.co/cerebras/Cerebras-GPT-13B to LLaMA-13B https://github.com/facebookresearch/llama/blob/main/MODEL_CA... you can see that in all of
125.
by amilios 4y ago
We don't know, OpenAI refused to publish any details about the architecture in the technical report. We don't know parameters, we don't know depth, we don't know how exactly it's integrating image data (ViT-style ma
126.
by amilios 4y ago
This is mostly true, yeah. Although in certain cases OpenAI is indeed pushing research forward themselves (not only using existing algorithms/architectures just at scale) and therefore has first mover's advantage, e.g. with reinfo
127.
by amilios 4y ago
OK sure life won't cease as a whole for humans, but living standards for the vast majority of us will enter freefall and stabilize at a MUCH much lower level than we're used to.
128.
by amilios 4y ago
I think the issue is feedback loops. You're right that eventually governments will be "forced" to do the "right thing", but the problem is that it is likely going to be too little too late, and the climate feedback
129.
by amilios 4y ago
This is an absolute myth. There is no evidence of "immunity debt", and serious public officials parrotting this nonsense are doing real damage. It's a tempting explanation because it's simple and seems to make intuitive
130.
by amilios 4y ago
Unfortunately I'm quite sure it's the latter. Even in Canada, in MTL where I live, on public transit it's maybe 5-6% that is masking. For most people the pandemic is simply "over", regardless of what the data says o
131.
by amilios 4y ago
It's not, not in the ChatGPT web interface at least. It's just a chat prompt, that's it. Can't control any sampling parameters at all, not temperature, nor top-P or top-K-based selection, nothing. I guess OpenAI is aimin
132.
by amilios 4y ago
In the same way that it predicts a given token like 'we' or 'write', it predicts a special end-output token, something analogous to 'EOF'. So at some point as it's generating new tokens and coming up with
133.
by amilios 4y ago
The "knowing what it doesn't know" problem is trickier even than it might seem, too: right now we don't even know how what it does know is encoded in its neurons, other than some vague notion that the knowledge is some
134.
by amilios 4y ago
The concern with Ticketmaster is the monopoly though, you are sidestepping the issue. It's about what kind of value Ticketmaster adds (minimal) to justify the exhorbitant fees they charge and can charge, because they have no competitio
135.
by amilios 4y ago
Sometimes I think I'm good at programming, then I find something like this and I'm put back in my place.
136.
by amilios 4y ago
It's written, it just doesn't have a standardized written form, it's written informally in standard Chinese character but it's modified from the spoken form in order to conform better to standard Mandarin forms, AFAIK. F
137.
by amilios 4y ago
The article says that the Amazon recommendation algorithm was also recommending along with the sodium nitrite, "a scale to measure a lethal dose, a drug to prevent vomiting, and a book with instructions on how to use the chemical to at
138.
by amilios 4y ago
There's a significant quality difference however if you look at the generated samples in the paper. Imagen Video is leagues ahead. The progress is still quite drastic
139.
by amilios 4y ago
There's actually a dating app with a similar-ish concept called Happn ( https://happn.com/ ), which basically is Tinder but if Tinder only showed you the profiles of people you had actually crossed paths with in real lif
140.
by amilios 4y ago
Online courses? Do you mean "investing in yourself" per se? Any ideas for growing the money that involve actual investment instruments, if I hypothetically have $10k lying around?
141.
by amilios 4y ago
Because I am not a US citizen :)
142.
by amilios 4y ago
The ballotpedia summary shows that there wasn't really a campaign opposing the measure it seems? In which case presumably it makes sense that it passed overwhelmingly, if no one was voicing any arguments against it
143.
by amilios 4y ago
What is the biggest Ada-based project currently in use? Does anyone know? Is Ada more for DoD-type backend-stuff, where Fortran or C would be used, or something else? What is Ada for?
144.
by amilios 4y ago
For anyone looking for schools for CS, https://csrankings.org/ is an excellent site that gives you an overview of departments' research output, filterable by subfield, geographic location etc. A very nice system if you
145.
by amilios 4y ago
On the flipside, natural language interfaces have the potential to be extremely easy to use for anyone, including non-experts. Anyone can type a message to the computer, without having to learn the specifics of an interface's custom co
146.
by amilios 4y ago
How exactly do game engines monetize, can anyone shed any light? Is it via additional support in using the engine for large companies? A freemium model like Unity's? But didn't even Unity succumb eventually to supplementing moneti
147.
by amilios 4y ago
It's interesting, because through my readings in linguistics, I got the sense that no language is "objectively" harder than any other, that everything is relative to the other languages you know. A Dutch speaker will learn Ge
148.
by amilios 4y ago
Is there any point on which Tensorflow actually wins out against Pytorch? My sense is that Pytorch is easier for almost every if not every single use case. Not to mention the 1 => 2 transition gave me serious Python 2 => 3 vibes with
149.
by amilios 4y ago
How long does it take to generate a single image? Is it in the 30 min type range or a few mins? It's hypothetically "possible" to run e.g. OPT175B on a consumer GPU via Huggingface Accelerate, but in practice it takes like 30
150.
by amilios 4y ago
Oh shoot, I thought it was for everything. Fair enough. Although I think new competitive features like Outpainting are definitely in response to Stable Diff

More ›